AI agents are just the start. Here’s what comes next

AI agents won’t transform your business alone—but with the right strategy, they can deliver real impact. As the artificial intelligence race accelerates, IBM’s Francesco Brenna argues that success lies not in the tools themselves, but in rethinking how work gets done—and why IBM’s agentic app model could be the real game-changer

There’s a lot of noise out there when it comes to AI agents. Some of it’s justified – the technology is moving fast – but much of the conversation is still fuelled by hype rather than practical progress. The truth is, while AI agents are already making a meaningful impact in the workplace, they’re not the fully autonomous problem-solvers some vendors make them out to be. Not yet, anyway.

At IBM Consulting, we’re seeing growing interest from organisations wanting to understand how AI agents can fit into their business. These agents are essentially autonomous bits of software that can carry out a range of tasks—analysing data, assessing situations, solving problems—with minimal human input. But on their own, each agent is limited. To be truly effective, they need to work together as part of an orchestrated network, each one playing its part in more complex workflows.

That said, there are limits. AI agents still require structure, oversight, and clearly defined goals. Businesses can’t simply ‘plug them in’ and expect them to deliver perfect outcomes. Human intervention is still needed to validate outputs and to ensure what’s being produced aligns with wider strategic and ethical priorities.

The bigger challenge, in my experience, is less about what the AI agents can or can’t do, and more about how organisations are trying to use them. Too often, businesses are looking to layer AI on top of old systems and outdated ways of working. The result? Limited gains.

Legacy IT stacks, with their mix of disconnected applications and outdated integrations, create serious hurdles. Data is fragmented, inconsistent, and often simply not fit for purpose. Without reliable data and modern processes, even the most advanced AI can only go so far.

This is why we’ve been talking to clients about something we call ‘agentic’ apps. These are purpose-built business applications that bring together multiple AI agents in a structured, networked way to handle end-to-end workflows. But—and this is key—it’s not just about the tech. It’s about rethinking how work gets done.

To make AI agents deliver real value, three core layers need to come together. First, we need to look at how people interact with AI. This starts with the front end, where AI assistants are tailored to different employee roles—what we call user personas—so the technology slots naturally into the tools people are already using. Done right, this substantially enhances both usability and productivity.

Second is the orchestration itself. This is the engine room, where instead of AI agents working in silos, we build systems where they operate as part of an interconnected network. Each agent handles a specialised task, but together they manage broader, more complex processes. It’s this kind of coordination that turns a series of narrow capabilities into something more powerful.

And finally, there’s the data. AI can only ever be as good as the data it runs on. That means ensuring data is clean, consistent, and accessible across the enterprise in real time. It sounds obvious, but it’s often the hardest bit to get right.

Adoption is happening, but it’s not uniform—and it’s not overnight. Most companies are starting in areas where they can see clear return on investment. We’re seeing good progress in customer service, data handling, IT operations, and software development.

Take software engineering, for example. AI agents are already helping developers with repetitive tasks—refactoring code, suggesting improvements, identifying bugs. An agent might spot recurring issues in a codebase, propose efficiency tweaks, or even offer better algorithms. But even here, the human developer is still in charge. They decide which changes to accept, how the design should evolve, and what the final product looks like.

So we’re not replacing developers. We’re giving them more time to focus on creative and high-value work. And that’s a pattern we’re seeing across industries. AI agents are augmenting roles, not replacing them.

That shift—from automation to augmentation—brings with it a new set of demands. If organisations want to get the best out of AI agents, they need people who know how to work with them. That means training teams to guide, validate, and optimise what the AI delivers.

Engineers and analysts will need to get comfortable not just using AI tools, but shaping them—setting their boundaries, monitoring their output, and ensuring they’re aligned with business goals. The workforce of the future will need to be as fluent in AI as it is in spreadsheets or CRM systems.

If there’s one message I’d underline, it’s this: AI agents are not a silver bullet. You can’t just drop them into a business and hope for transformation. What’s needed is a broader rethink—a digital transformation strategy that includes AI, but also covers process redesign, data governance, team capability, and strong oversight.

We call this discipline ‘AI operations’—the business of managing, optimising, and governing how AI agents are deployed. It’s where trust, explainability, and interoperability come into play. And it’s how organisations can ensure that human-AI collaboration is not just possible, but productive.

There’s no question that AI agents will become a major force in the modern workplace. But the biggest wins will go to those organisations that take a thoughtful, strategic approach. That means looking beyond the hype, getting the foundations right, and staying focused on where real value can be delivered.

When used well, AI agents will help people to work faster and smarter. As AI evolves, the organisations that rethink how work gets done will be the ones that stay ahead.